[libc-commits] [PATCH] D141997: [libc][NFC] Detect host CPU features using try_compile instead of try_run.

[Guillaume Chatelet via Phabricator via libc-commits]

gchatelet added a comment.

Thx for the patch @sivachandra! It's definitely better than what I did, I was unhappy with the design but it was a first step.

There are still a few issues in my opinion:

- in msvc features have a different name <https://learn.microsoft.com/en-us/cpp/preprocessor/predefined-macros?view=msvc-170>, this means that the way we refer to features in CMake cannot be the same as what is put into the file,
- the way we detect features in CMake and in llvm libc can be different,
- what is actually compiled by CMake is hard to understand.

I think this can be improved further.

Here is what I have in mind :

- Add preprocessor feature detection to something like `libc/src/__support/cpu_features.h` (we already have `LIBC_TARGET_HAS_FMA` in `libc/src/__support/architectures.h`). This file would only depend on freestanding headers like `compiler_features.h` or `architectures.h` (sidenote we should probably make it clear in the filename that these headers are freestanding / only preprocessor)
- Create one cpp file per feature with a filename similar to the LIBC preprocessor definition. We don't have a lot of features so the number of files will stay low. This file would use the libc header so we have consistent detection between CMake (Bazel?) and the code.

WDYT? I can take a look at it if you want?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D141997/new/

https://reviews.llvm.org/D141997